The corneal endothelium (CE) is a monolayer of cells localised in the innermost segment of the cornea, regulating solutes transport from and to the aqueous humor (pump-leak hypothesis)1. Corneal endothelial cells (CEnC) are generally considered non-dividing in vivo2 and arrested in the G0/G1 phase of the cell cycle3,4. The presence of anti-proliferative factors in the aqueous humor, the stress-induced premature senescence, and the contact inhibition between cells are the main events triggering the expression of cell cycle regulators that induce CEnC mitotic block2. For this reason, as CEnC density decreases by 0.6% each year5,6, cell loss is compensated through migration and cell enlargement2,7. When, as a consequence of pathologies or surgical treatments, endothelial cell density falls below 500 cells/mm2, cell enlargement is no longer able to compensate for the passive leaking7. In this case, the impaired CE results in corneal swelling and, if untreated, in the subsequent severe loss of corneal transparency, ultimately leading to blindness.
CEnC dysfunction is the underlying cause of about 40% of all corneal transplants performed worldwide7. Corneal grafts, despite decisive advancements in the last decades, are still complex and invasive procedures, with some severe limitations, including immune graft rejection, graft failure, and a general scarcity of donor corneas7.
Alternative clinical procedures have been recently proposed in order to increase the number of patients that can be treated and to improve their quality of life. Currently, descemetorhexis is the only therapeutic approach that avoids allogeneic transplantation. This treatment involves the removal of the central CEnC to directly stimulate migration of the patient’s own peripheral CEnC towards the central cornea. Unfortunately, although promising, the procedure can be applied only in those cases that have preserved a high density of CEnC in the peripheral cornea (i.e., in some cases of Fuchs endothelial dystrophy), and the rate of success is still low, requiring a long time for the post-operative recovery7.
Interestingly, successful studies have recently promoted the development of a cell-based therapy8,9, which may be considered as a less invasive substitute for a corneal graft. Since induced mitosis has been observed in cultured CEnC10, heterologous cell transplantation to the patient’s anterior chamber turned out to be a promising strategy. However, despite CEnC expansion in vitro has been thoroughly investigated11,12, only in 2018 Kinoshita’s group demonstrated for the first time that injected CEnC from young donors are able to repopulate the cornea of a cohort of patients8. Two factors, in particular, have made CEnC culture in vitro very challenging: their reduced proliferative capacity, exacerbated over passages by replicative senescence12, and the propensity to spontaneously lose their morphology through an endothelial-to-mesenchymal transition (EnMT)13.
For implementation and efficient substitution of corneal graft in clinical practice, these or other new approaches need to be investigated. First, it is necessary to acquire deeper knowledge on how to regulate and circumvent the mitotic block. One prominent approach aiming to reach this goal was focused on the examination of the role of several growth factors. Among them, basic fibroblast growth factor (bFGF)14,15,16, insulin-like growth factor (IGF)16 and hepatocyte growth factor (HGF)16, all present in the aqueous humor, stimulate CEnC proliferation, while transforming growth factor beta (TGF-β), which is also found in the aqueous humor, induces CEnC mitotic block17,18. Intracellular pathways, activated by these growth factors, are shown not to act in parallel but rather to build a network of cross-regulation: bFGF promotes cellular proliferation in CEnC through PI3K/Akt4,19,20,21, while its activity is counteracted by the presence of TGF-β19,22. Nevertheless, in CEnC, the complexity of pathway regulation is not limited to the cross-talk between them but also implies the alternative functions, which the same pathway can overtake in different cellular conditions. For instance, TGF-β activation has been linked either to CEnC maturation or to EnMT23, while bFGF, although promoting proliferation19, was also found to be involved in EnMT24,25. Intriguingly, canonical Wnt signalling and consequent β-catenin nuclear internalization, triggered by bFGF together with EDTA-mediated cell junction disruption, promotes both an increase in CEnC proliferation and in EnMT26. In accordance with the aforementioned study, β-catenin in CEnC was described to promote cellular proliferation through the expression of cyclin D127, but also to regulate EnMT28. Taken together, these elements suggest that Wnt/β-catenin pathway may promote proliferation and, in parallel, regulate a sensitive balance between the maintenance of the polygonal morphology and the transformation to the fibroblastic phenotype (EnMT). Any variation in mechanotransduction, medium components or donor’s cell characteristics can impair this delicate equilibrium and induce development of one fate rather than the other.
Nevertheless, the role of this and other factors orchestrating CEnC proliferation while maintaining the correct morphology has to be further elucidated.
Similar to recent studies on novel eye surface cellular mechanisms29,30, we used here a hypothesis-free approach to compare the proteome of rabbit CEnC (rCEnC) isolated from the corneal tissue with the proteome of rCEnC grown in culture. This analysis allowed the identification of 3,328 rabbit proteins, 77 of which were differentially regulated at the two conditions with a p < 0.005. Several of these proteins were found to be associated with common intracellular pathways, in particular Wnt/β-catenin and TGF-β, which were further investigated in cultured rCEnC. Nuclear translocation of β-catenin was observed in mitotic cells while TGF-β was shown to arrest cells in the G2/M phase of the cell cycle as well as block β-catenin nuclear translocation. Moreover, β-catenin pathway inhibition significantly decreased cellular proliferation. By contrast, we did not observe any pro-proliferative effect upon direct Wnt pathway activation. In summary, our results suggested that β-catenin is essential, but not sufficient to overcome the mitotic block. In parallel, we confirmed that bFGF has a pro-proliferative effect on CEnC and that its activity is mediated by β-catenin activation and counteracted by TGF-β stimulation. Overall, the results provide new insights into the role of β-catenin in corneal endothelial proliferation in vitro, after cell–cell disruption and growth factors stimulation (bFGF and TGF-β).
Identification of cell cycle phases characterising rabbit CEnC in vitro and ex vivo
rCEnC growing in vitro were compared with ex vivo rCEnC directly isolated from the Descemet’s membrane, using flow cytometry and PI staining, in order to assess the cell cycle status of rCEnC in the experimental samples. Rabbit CE ex vivo exhibited a unique defined peak, representing cells in the G0/G1 phase of the cell cycle (Fig. 1a). rCEnC expanded in vitro were distributed instead among different phases of the cell cycle, depending on cell culture confluence: at 60% confluence, 20% of the population was in G2/M phases of the cell cycle (Fig. 1b), while confluent cells exhibited prevalence of the cells in G0/G1, similarly to those isolated from the tissue (Fig. 1c). Human CEnC showed a similar behaviour: all of them were found to be in the G0/G1 phase when isolated from the tissue and directly analysed (Fig. 1d), while they were redistributed in the three different phases with percentages similar to the rCEnC, when harvested at 60% confluence (Fig. 1e). Quantifying cell cycle distribution of both human and rabbit CEnC at 60% confluence we confirmed that human and rabbit CEnC, under the same culture condition, have a comparable rate of proliferating cells and indicated that the rabbit CE is an appropriate model for studying cell cycle regulation (Fig. 1f). Interestingly, human CEnC at higher passages (P4), presenting with an elongated morphology13, showed a significantly increased population of cells in G2/M phase compared to the previously analysed human CEnC (P2) (Supplementary Figure S1).
Proteomic comparison of rabbit CE and cultured rCEnC was performed in order to elucidate differentially regulated pathways between tissue and actively proliferating cultured cells. The proteomic analysis, carried out in a total of 30 samples, identified 3,328 proteins from the complete rabbit 20,190,508 database. The expression levels of these proteins were highly reproducible within the biological replicates under examination. This allowed the selection of proteins with a significant difference between the tissue and the cultured rCEnC: 77 proteins with a p value < 0.005, 19 of which with a p value < 0.001 (heat map in Supplementary Figure S2). Among those with a p value < 0.005, 16 proteins were found to be upregulated, while 61 proteins were downregulated in in vitro rCEnC as compared to the ex vivo rCEnC from the tissue (Tables 1, 2). A full protein dataset is available in supplementary material (Supplementary Table S1).
Gene ontology and literature analysis
Gene ontology analysis (Table 3) of the proteomic results revealed that some specific biological processes were significantly dysregulated, especially those involved in the formation of the proteasome complex, in the negative regulation of canonical Wnt signalling pathway, in the endoplasmic reticulum function, and in cellular trafficking (GO:0000502, GO:0090090, GO:0005789, GO:0006888, GO:0031201, GO:0031032 and KEGG pathway 04141). Other processes found in the gene ontology analysis, listed in Table 3, were not significantly modulated (p > 0.05).
Literature analysis showed that some of these proteins have been previously associated with CE function and dysfunction (Supplementary Table S2) and that several of them have a role in multiple fundamental intracellular pathways (Table 4). More specifically, based on the proteomic results, we found numerous proteins differentially expressed in proliferating in vitro versus ex vivo cells, which were involved in β-catenin and TGF-β pathways.
Several proteins involved in promoting the Wnt/β-catenin pathway were significantly downregulated in cultured rCEnC when compared with the tissue-derived counterpart. For instance, HIPK1 was previously shown to induce Wnt/β-catenin signalling31. KHDRBS1 promotes proliferation through the same pathway32, while ITGB5 inhibits β-catenin degradation by the proteasome, leading to Wnt/β-catenin pathway activation as well33. Similarly, Endoglin (ENG)34, CUX35, CARF36, SPARCL137, and SPARC38 were proved to trigger the Wnt/β-catenin pathway. Downregulation of proteins 4.1R (EPB41) and 4.1G (EPB41L2), which associates β-catenin to the cell membrane in gastric epithelium39, may play a role in cytoplasmic β-catenin translocation. Moreover, DAVID analysis predicted a negative regulation of the canonical Wnt signalling pathway (Table 3), identifying some proteasome proteins that were upregulated in rCEnC (PSMD7, PSMD2, PSMD12, PSMD3, PSMD6, and PSMB7). The PSMD are components of the 19S subcomplex, known to be involved in β-catenin/Wnt signalling; PSMD7 and PSMD2 in particular are found to be directly associated with the degradation of β-catenin from the proteasome40.
Although Wnt/β-catenin is involved in CE development41 and in CE mesenchymal transformation26,28,42, the mechanism of β-catenin pathway downregulation observed here is not completely clear and it will be further dissected in the next sections.
Multiple proteins that were found downregulated in cultured rCEnC are also involved in TGF-β pathway regulation. Endoglin, previously shown to be expressed in CEnC43, is an auxiliary receptor for TGF-β, controlling proliferation and quiescence44, in particular in the endothelium45, where it also regulates the EMT46. CUX1, a protein with a role in regulating cell cycle progression47, EMT48, and repressing E-cadherin49, is proved to be a target of TGF-β pathway47 and PSPC1 potentiates TGF-β autocrine signalling50. Similarly, expression of LRP1, a cell receptor involved in the clearance of growth factors, including TGF-β51 is increased upon stimulation of TGF-β151, and HPIP is a downstream factor of TGF-β1 stimulation, promoting EMT in A549 cells52.
These results show that Wnt/β-catenin modulators and TGF-β effectors are downregulated in culture-derived proliferating rCEnC and, in accordance with previous reports, demonstrate a fine-tuned regulation involved in promoting CEnC proliferation while still suppressing the EnMT.
β-Catenin cellular localization
Staining of rCEnC showed that β-catenin is mainly located in the plasma membrane when cultured cells are confluent and reconstitute the tissue-like alveolar structure (Fig. 2a, panel 1). Nevertheless, the external cells of the colonies, not presenting any cell–cell contact, do not show β-catenin on the plasma membrane (Fig. 2a, panel 2). Most importantly, cytoplasmic and nuclear translocation of β-catenin from the plasma membrane was observed in the dividing or in some isolated cells (Fig. 2a, panels 3–5). These data, together with available literature, suggest a complex mechanism containing two phases. In the first phase, β-catenin is released from the cell-junctions, moving eventually to the nucleus for regulation of the gene expression. In the second phase, it is degraded or transferred back to the membrane, while its activators are downregulated, as observed in the proteomic results (Fig. 2b,c).
We suggest this hypothesis based on the observation of β-catenin translocation in the first phase, which can also explain why it was quickly degraded and subsequently downregulated.
Variability in the cell cycle and β-catenin distribution after treatments with bFGF and TGF-β
In order to increase or diminish the number of cells in the G2/M phase of the cell cycle and analyse in parallel the effect on β-catenin nuclear translocation as well as the maintenance of a corneal endothelial phenotype, we treated rCEnC with different compounds.
Initially, bFGF was used as a positive control for proliferation induction in CEnC19 and its cross-regulations with the Wnt pathway, as observed in other cellular models53 and in CEnC26. Moreover, bFGF signalling was shown to downregulate GSK-3β activity through a mechanism involving Akt, which in turn activates β-catenin response54. As expected, rCEnC treated with 20 ng/mL of bFGF showed a significant reduction in the number of non-proliferating cells and a consequent increase in the G2/M population (Fig. 3a).
In parallel, TGF-β was tested on rCEnC as it correlates with multiple proteins that emerged from the proteomic analysis and has an ascertained role in reducing CEnC proliferation17,18,55,56,57,58. Herein, TGF-β-treated cells presented a significant decrease in G2/M phase of the cell cycle as it was detected by using cytofluorimetric analysis, confirming the anti-proliferative effect of TGF-β on CEnC. When TGF-β was added to bFGF-stimulated cells, we observed a distribution comparable to the TGF-β alone treatment, proving the TGF-β interference with the bFGF-activated pathway (Fig. 3a).
The same treatments were analysed for β-catenin expression by immunocytochemistry. We detected a significant increase of β-catenin nuclear translocation in bFGF-treated cells, while the cytoplasmic signal was significantly reduced both in TGF-β and TGF-β + bFGF-treated rCEnC when compared to the control (Fig. 3b). These data collectively proved that the activity of bFGF was mediated by an increased translocation of β-catenin to the nucleus, which is inhibited by TGF-β, in accordance with the previous literature19.
The studies by immunofluorescence on α-SMA (Fig. 4), a marker of EMT, did not show any significant evidence of mesenchymal transformation. The amount of α-SMA-positive cells was not modified by bFGF treatment, as similarly assessed by Tseng and Heur groups26,59. Although TGF-β can induce EnMT42,60 and differently from what observed by Tseng et al.26, the treatment of rCEnC with either TGF-β and TGF-β + bFGF did not show any significant increase in α-SMA expression. Nevertheless, in the Tseng et al. study, the treatment lasted 3 days and was followed by EDTA dissociation, likely provoking a more sustained stimulation of β-catenin as well as other cellular responses.
The data obtained with α-SMA were corroborated by immunostaining with an early marker of EMT, S100A461,62, expressed within the cytoplasm by human adult CEnC in vivo63. Conversely, S100A4 expression was observed in the nucleus when CEnC underwent EnMT26. Similarly to what observed for α-SMA, we did not detect any significant variation in S100A4 expression between the treatments with bFGF, TGF-β, TGF-β + bFGF and the Mock control (Fig. 4a,b). In each treatment tested S100A4 presented as mainly cytoplasmic and/or perinuclear. As a positive control we used a rCEnC strain with a high number of passages (P10) which showed an elongated phenotype. In this condition S100A4 was localized in the nuclei of the majority of rCEnC, which were also showing a high α-SMA positivity (Fig. 4c). This result confirm that both proteins may be considered as valid markers for EnMT.
Altogether, these results showed that bFGF and TGF-β treatments did not cause any mesenchymal transformation on rCEnC at 24 h, although they were able to interfere with β-catenin and activate or inhibit proliferation. Further experiments in the following section, using small molecules targeting β-catenin pathways, helped to reveal a possible role of this crosstalk in rCEnC maintenance and propagation.
Variation in the cell cycle phases and β-catenin distribution after treatments with Wnt activators/inhibitors
CHIR99021 was previously described to inhibit GSK-3β64, thereby stabilizing cytoplasmic β-catenin and eventually promoting its nuclear translocation. On the basis of CHIR99021 IC50 (0.04 µM)64, the treatment was initially tested in a range of concentrations between 0.05 and 10 µM. The distribution in the cell cycle phases was not statistically different for all the CHIR99021 concentrations tested except for 10 µM. This concentration produced a significant decrease of cells in the G2/M phase of the cell cycle in comparison with untreated cells (Fig. 5a), although promoting a consistent β-catenin nuclear translocation (Fig. 5b,c). Interestingly, CHIR99021 at 0.5 µM, despite not showing any significant difference in cell phases distribution, revealed an increase in cytoplasmic and nuclear β-catenin if compared with the untreated cells (Fig. 5b,c). Collectively these results suggest that CHIR99021, although able to promote β-catenin migration to the nuclei, did not cause an increase in rCEnC proliferation. Conversely, at high concentration (10 µM), CHIR99021 decreased rCEnC proliferation. This unexpected effect might be due to a negative feedback regulation of β-catenin, once overactivated. The possibility of β-catenin feedback regulation was also previously proposed by Hirata‐Tominaga et al.28. CHIR99021 concentrations of 0.5 and 10 µM were also tested at 12 and 48 h: while at 12 h both concentrations did not show any effect on the cell cycle of rCEnC, at 48 h only 10 µM CHIR99021 elicits an effect not significantly different to what observed at 24 h (data not shown). Moreover, a vitality assay using calcein AM/Propidium Iodide staining was performed to assess if CHIR99021, at the concentrations used for the experiments on rCEnC, demonstrated a cytotoxic effect. However, we could not observe any Propidium Iodide positive cell nuclei, confirming that CHIR99021 and DMSO, at all the concentration tested under the same experimental conditions used in this study, did not induce cell death (Supplementary Figure S3).
In addition, we observed an increased amount of α-SMA positive cells and nuclear S100A4 at 10 µM CHIR99021 (Fig. 5b,c), suggesting that activation of β-catenin nuclear translocation over a certain limit may induce EnMT. Similar results were previously obtained via disrupting CEnC junctions with EDTA and with prolonged addition of TGF-β (2 days)26. Under these conditions, rCEnC might thus interfere with the β-catenin pathway to avoid the loss of hexagonal morphology and the mesenchymal transformation.
Next, we tested an inhibitor of β-catenin pathway: quercetin65,66, previously proven to inhibit the Wnt/β-catenin signalling by interfering with β-catenin nuclear translocation67. The 25 µM concentration was selected for our study. Quercetin was shown here to completely abolish the rCEnC populations both in S and G2/M phases, maintaining the cells in the G0/G1 phase of the cell cycle (Fig. 5d). Similar results were obtained by treating rCEnC with both bFGF and quercetin (Fig. 5d), highlighting how the latter was also able to knockdown the proliferative effect of bFGF. These data suggest that quercetin may act, at least in part, via the same pathway activated by bFGF.
Overall, the presented results show how β-catenin appears necessary but not sufficient to promote proliferation of rCEnC. The β-catenin pathway is activated by the disruption of cell–cell junction and β-catenin expression is increased by bFGF, while it is inhibited by the presence of TGF-β. However, β-catenin is quickly degraded by cells, underlining the sophisticated regulation of this specific pathway in the corneal endothelium, with the likely effect to avoid EnMT.
Several studies have been carried out with the aim to characterize how CEnC regulate their proliferation, mainly focusing on the role of specific growth factors (EGF68, bFGF19, and TGF-β55,56) and the final downstream effectors, in particular those involved in the regulation of the cell cycle (cyclins, p16, p21, and p272,57). Herein we propose a hypothesis-free approach, starting directly from the analysis of the CE proteome through a correlation between ex vivo non-proliferating rCEnC and in vitro proliferating rCEnC. Using rabbits farmed in standard conditions (controlled environment, slaughtered at the same age) and correlating the two eyes of the same individual, we were able to minimize the inter-individual variations. Moreover, although cellular division was observed in rabbit CE in vivo69, human and rabbit CEnC showed a similar cell cycle distribution, either when isolated from the tissue or when cultured in vitro. For this reason, rCEnC were selected as a model to study proliferative mechanisms that are utilised by the cells when expanded in vitro. The proteomic analysis allowed the selection of 77 proteins, out of the 3,328 identified, with a significant differential expression between the two groups in comparison (p < 0.005). A targeted dissection of each identified protein and their related intracellular routes, enabled to propose a correlation between specific pathways and rCEnC expansion in vitro.
For instance, several proteins promoting activation of the Wnt/β-catenin pathway, including HIPK131,70, KHDRBS132, ITGB533, Endoglin (ENG)34, CUX35, CARF36, SPARCL137, and SPARC38, were all significantly reduced in cultured rCEnC. Negative regulation of the canonical Wnt signalling pathway was further confirmed by upregulation of some proteasome proteins (PSMD7, PSMD2, PSMD12, PSMD3, PSMD6, and PSMB7; Table 3)40.
Beside Wnt/β-catenin, multiple proteins involved in the TGF-β signalling were found to be downregulated. In particular, the expression of TGF-β effectors like Endoglin44,45,46, CUX147,48,49, LRP151, HPIP52, and PSPC150 was significantly reduced in rCEnC.
Taken together, the proteomic characterization suggested that Wnt/β-catenin and TGF-β might have a key function in proliferating rCEnC (Table 4) and garnered therefore an increasing attention in this study.
Proteins of the TGF-β family are cytokines activated upon different stimulation (i.e., injury or mechanical stress)5. In the cornea, TGF-βs (1, -2, and -3) are present in the aqueous humor71 and, once released, trigger multiple downstream processes72. A double role was recently described in CEnC for TGF-β: it induces the correct morphology and formation of novel cell junctions during the maturation phase, while it promotes a fibroblastic phenotype during proliferation23. This double role reflects its already observed activity in other studies, where TGF-βs is involved in the correct maturation of CE41, in stimulating CEnC migration during wound healing26, in promoting EnMT42,60 and, most importantly, in suppressing CEnC entry into the S-phase17,18,55,56,57,58.
Following the proteomic results, downregulation of some TGF-β effectors observed in dividing rCEnC may be consequent to the absence of a TGF-β stimulus in vitro: cultured rCEnC, no longer blocked in a non-proliferative state by TGF-β, acquire in this manner the ability to proliferate.
The other pathway arising from the proteomic analysis involves Wnt/β-catenin signalling. When the cells are in contact with each other, β-catenin is usually bound to the cell membrane in complex with cadherin (N-cadherin in CE73) while it is released when cells lose their cell–cell junctions74. Once in the cytoplasm, free β-catenin is targeted for degradation by GSK-3β phosphorylation. However, when Wnt activates the canonical pathway, it promotes the stabilisation of the cytoplasmic β-catenin and eventually, its translocation into the nucleus5, where it drives transcription of target genes, in particular of those that are critical in promoting cell proliferation, such as c-myc and cyclin D175. Proliferation is known to be induced by β-catenin activation in different cell types, including cancer76 and stem cells74,77. In CEnC, β-catenin acts as a key regulator of two crucial mechanisms: proliferation and cellular morphology26,27 but also in expressing CE markers, as observed following treatment with a GSK-3β inhibitor (6-bromoindirubin-3′-oxime, BIO) which increases β-catenin levels41.
Herein we attempt to explain the role of β-catenin during rCEnC proliferation and uncover the reasons of its rapid degradation in the cytoplasm via cellular machinery. In fact, we observed binding of β-catenin to the cell membrane in confluent cells in vitro, while it disappeared from the membrane when the cell–cell junctions were lost (Fig. 1). This result was confirmed by the proteomic characterization that indicated activation of proteasome components, previously linked to β-catenin degradation40 (Tables 1, 2, 3). Moreover, the majority of cells in culture were shown to downregulate various proteins involved in β-catenin activation. Of those proteins, Endoglin43, ITGB578, and SPARC79 were previously found to be expressed in CEnC (Supplementary Table S2). β-catenin is indeed generally fast degraded once in the cytoplasm75, and the condition captured by the proteomic analysis showed how β-catenin may have been already digested by the proteasome or moving back to the cell membrane (Fig. 2b,c). At the same experimental condition, half of rCEnC population (60% confluence) was in the G0/G1 phase, while only a minority of them were actively proliferating in the G2/M phase of the cell cycle (Fig. 1b). However, within this population, we could observe a nuclear β-catenin localisation in isolated or mitotic CEnC (Fig. 2a), suggesting the importance of this translocation during cell division.
Collectively, these results raised questions regarding the possible role of β-catenin in rCEnC expansion that we tackled using a GSK-3 inhibitor, CHIR99021, to induce β-catenin stability, and quercetin as a counteracting inhibitor of β-catenin nuclear translocation67. Treating rCEnC with quercetin, we observed a substantial decrease in the cells in the G2/M phase of the cell cycle (Fig. 5d), while we never found a corresponding increased amount of cells in the G2/M phase at all tested CHIR99021 concentrations (Fig. 5a). Conversely, CHIR99021 (at 10 and 0.5 µM) was able to raise drastically the number of cells where β-catenin had translocated to the nucleus (Fig. 5b,c). Taken together, these data confirm how β-catenin expression and localization are fundamental during cell proliferation process, in accordance with previous reports26,27, but probably not sufficient to promote it.
Based on the described role of β-catenin in dividing rCEnC and on the effect of growth factors in eliciting in vitro CEnC proliferation19, we sought to understand whether specific growth factors would act via β-catenin activation. β-catenin was shown to be involved in orchestrating the downstream activity of FGF in various cell types54,80, and, as previously introduced, bFGF stimulates CEnC proliferation through PI3K/Akt activation4,19,20,21, while TGF-β19,22 inhibits cell proliferation elicited by bFGF54. Here we investigated how the pro- and anti-proliferative activities of these two specific growth factors might crosstalk through β-catenin signalling. Consistently with previous reports, we confirmed the role of bFGF in promoting CEnC proliferation19,26 showing how bFGF mediates an increase in the G2/M phase of the cell cycle and in β-catenin nuclear translocation (Fig. 3). Moreover, bFGF pro-proliferative effect was completely abolished by the quercetin-mediated inhibition of the β-catenin pathway (Fig. 5d). Similarly, we assessed the anti-proliferative effect of TGF-β17,18,55,56, which exerted an analogous decrease in the number of cells in the G2/M phase either alone or in the presence of bFGF, thus abolishing any pro-proliferative effect of the latter (Fig. 3a). We proved for the first time here that this effect, described also by Lu and collaborators19, is mediated by the inhibition of β-catenin translocation to the nucleus (Fig. 3b). Collectively, the results reported in our study confirmed the importance of β-catenin during CEnC proliferation process both after cell–cell disruption and bFGF stimulation. Consistent with this finding and with previous reports, TGF-β inhibits proliferation by blocking β-catenin nuclear translocation.
As introduced earlier, the proliferation process is often balanced in a delicate equilibrium with EnMT. Although more specific experiments are necessary to clarify the role of β-catenin in EnMT, we measured a significant increase in the number of α-SMA and S100A4 (markers of EMT) positive CEnC when treating cells with 10 µM CHIR99021 (Fig. 5b,c). Such evidence suggests that a sustained stimulation with CHIR99021 may induce β-catenin nuclear translocation and that EnMT ensues when β-catenin is activated over a certain threshold. Similarly, Zhu et al. confirmed how CEnC loosing cell–cell junctions after treatment with EDTA exhibit a defined nuclear localization of β-catenin, promoting proliferation but also EnMT26, whereas Kinoshita et al. observed that CEnC attempt to block EnMT by deregulating β-catenin28. Altogether these results confirmed that β-catenin expression requires fine regulation since a sustained stimulus towards β-catenin activation might induce CEnC transformation into a mesenchymal phenotype, similar to cancer progression75. This observation may also explain why CEnC degrade β-catenin in the cytoplasm and deregulate β-catenin-activating pathways immediately after cell division, enabling thereby a putative CEnC mechanism of protection from β-catenin induced EnMT.
The data presented here suggest a scenario where β-catenin is necessary for CEnC proliferation, but its over-activation drives cells to EnMT. However, β-catenin alone is not sufficient to unlock the mitotic block requiring other cellular effectors, as those activated by the disruption of cell–cell membranes and stimulated by growth factors. For instance, RAC1 has been proposed to act in concert with β-catenin in inducing CEnC proliferation through Cyclin D1 transcriptional activation27.
In conclusion, herein we present a study that, starting from a proteomic analysis, provides insights into the major intracellular pathways that CEnC activate in culture. Further studies of these mechanisms are fundamental to understand how to unlock the CEnC mitotic block and improve their regenerative capacity in order to develop a localised therapy that would overcome the need for corneal transplantation.